Single Echo MRI Reconstruction Using Deep Learning

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Current MRI technologies face challenges in reducing acquisition time, RF power deposition, peripheral nerve stimulation, and gradient noise, particularly in single echo acquisitions, which often require multiple receive channels or external sensors.

Innovation Solution

The system employs a single echo reconstruction method that generates coil sensitivity weighted projections, inverts these projections using deep learning procedures, and concatenates them to reconstruct images efficiently, eliminating the need for external sensors and reducing noise, while utilizing a 64-channel coil without phase encoding.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Loss of time

If single echo acquisition is used to reduce acquisition time, then acquisition time is reduced, but image quality deteriorates due to blurring artifacts

Engineering Contradiction:

Improveacquisition timeVSAvoidimage quality

Core Design Contradiction:

Loss of timeVSManufacturing precision

Solution Approach 1:

The patent introduces deep learning algorithms as an intermediary between the single echo acquisition data and the final image reconstruction. The neural network processes the undersampled k-space data and learned coil sensitivities to generate high-quality images, acting as a mediator that transforms limited data into diagnostically useful images while suppressing blurring artifacts

Inventive Principle:

Principle #24Intermediary (Mediator)

Solution Approach 2:

The patent performs preliminary action by pre-calculating and storing coil sensitivity maps through training scans before the actual imaging process. These pre-computed sensitivity maps are then reused during rapid single echo acquisitions to accelerate reconstruction without compromising image quality, enabling the system to handle the reduced data from single echo sequences

Inventive Principle:

Principle #10Preliminary action

2Manufacturing precision

If multiple receive channels are used to improve image quality, then image quality is improved, but device complexity increases

Engineering Contradiction:

Improveimage qualityVSAvoiddevice complexity

Core Design Contradiction:

Manufacturing precisionVSDevice complexity

Solution Approach 1:

The patent implements self-service by enabling the coil array to automatically determine its own sensitivity distributions through training scans. The system uses the acquired data to learn and store coil sensitivity maps that are specific to each coil's characteristics, eliminating the need for manual calibration or external sensors while improving image quality through optimized signal combination

Inventive Principle:

Principle #25Self-service

3Loss of time

If external magnetosensors are used to enable single echo reconstruction, then single echo reconstruction is enabled, but device complexity and cost increase

Engineering Contradiction:

Improveacquisition timeVSAvoiddevice complexity

Core Design Contradiction:

Loss of timeVSDevice complexity

Solution Approach 1:

The patent extracts and eliminates the need for external magnetosensors by using only the existing MRI coil array. The system processes the signals from the standard receive coils through deep learning algorithms to achieve single echo reconstruction, removing the additional hardware components while maintaining the time-saving benefits

Inventive Principle:

Principle #2Taking out (Extraction)

Solution Approach 2:

The patent replaces the mechanical/physical approach of using external sensors with a computational approach using deep learning algorithms. Instead of adding physical sensing elements, the system uses software-based signal processing and neural networks to extract the necessary information from the existing coil signals, achieving the same goal with reduced hardware complexity

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

4Object-affected harmful factors

If RF power deposition is reduced to improve safety, then safety is improved, but signal quality deteriorates

Engineering Contradiction:

ImproveRF power depositionVSAvoidsignal quality

Core Design Contradiction:

Object-affected harmful factorsVSManufacturing precision

Solution Approach 1:

The patent changes the acquisition parameters by using single echo acquisition with reduced RF power deposition. By modifying the pulse sequence to use lower RF power and accepting the resulting data limitations, the system prioritizes patient safety while using deep learning reconstruction to compensate for the reduced signal quality

Inventive Principle:

Principle #35Parameter changes

Data Source

PatentUS20230326102A1System, method, and computer-accessible medium for facilitating single echo reconstruction of rapid magnetic resonance imaging

Publication Date: 2023.10.12 THE TRUSTEES OF COLUMBIA UNIV IN THE CITY OF NEW YORK

AI summary

An exemplary system, method, and computer-accessible medium for reconstructing a portion(s) of an image(s) of a patient(s) can include, for example, receiving magnetic resonance imaging (MRI) information for the patient(s), generating a plurality of coil sensitivity weighted projections based on the MRI information, inverting a column in the coil sensitivity weighted projections to generate inverted column information, and reconstructing the portion(s) of the image(s) based on the inverted column information. The portion(s) of the image(s) can be deblurred, for example, using a deep learning procedure(s). A reference scan of a part(s) of the patient(s) can be received, and deep learning procedure(s) can be trained based on the reference scan.